Hugging Face's logo

optimum-internal-testing
/
neuron-testing-cache

Model card Files
xet
 Community
optimum-internal-testing-user commited on
Commit
14db955
·
verified ·
1 Parent(s): c18a007

Synchronizing local compiler cache.

Browse files
Files changed (8) hide show
  1. neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev4/gemma3/unsloth/gemma-3-4b-it/bea7134934abbc294477.json +130 -0
  2. neuronxcc-2.21.33363.0+82129205/MODULE_15503648c89ebbfa6dc4+24129607/model.hlo_module.pb +1 -1
  3. neuronxcc-2.21.33363.0+82129205/MODULE_15503648c89ebbfa6dc4+24129607/model.neff +1 -1
  4. neuronxcc-2.21.33363.0+82129205/MODULE_2a8f4b110978c8fcb167+c4f887dc/model.hlo_module.pb +1 -1
  5. neuronxcc-2.21.33363.0+82129205/MODULE_2a8f4b110978c8fcb167+c4f887dc/model.neff +1 -1
  6. neuronxcc-2.21.33363.0+82129205/MODULE_2a8f4b110978c8fcb167+c4f887dc/wrapped_neff.hlo +1 -1
  7. neuronxcc-2.21.33363.0+82129205/MODULE_976a4227c74e1e5d858d+a02c3a36/model.neff +1 -1
  8. neuronxcc-2.21.33363.0+82129205/MODULE_976a4227c74e1e5d858d+a02c3a36/wrapped_neff.hlo +1 -1
neuronxcc-2.21.33363.0+82129205/0_REGISTRY/0.4.6.dev4/gemma3/unsloth/gemma-3-4b-it/bea7134934abbc294477.json ADDED
@@ -0,0 +1,130 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_entry_class": "SingleModelCacheEntry",
3
+ "_model_id": "unsloth/gemma-3-4b-it",
4
+ "_task": "image-text-to-text",
5
+ "architectures": [
6
+ "Gemma3ForConditionalGeneration"
7
+ ],
8
+ "boi_token_index": 255999,
9
+ "dtype": "bfloat16",
10
+ "eoi_token_index": 256000,
11
+ "image_token_index": 262144,
12
+ "initializer_range": 0.02,
13
+ "mm_tokens_per_image": 256,
14
+ "model_type": "gemma3",
15
+ "neuron": {
16
+ "_serialized_key": "NxDVLMNeuronConfig",
17
+ "batch_size": 2,
18
+ "capacity_factor": null,
19
+ "checkpoint_id": "unsloth/gemma-3-4b-it",
20
+ "checkpoint_revision": "bf46152c47f5dd20b896357cb51abc4c03b8ee8c",
21
+ "continuous_batching": true,
22
+ "ep_degree": 1,
23
+ "fused_qkv": true,
24
+ "glu_mlp": true,
25
+ "image_seq_len": 256,
26
+ "image_size": 896,
27
+ "local_ranks_size": 2,
28
+ "max_batch_size": 2,
29
+ "max_context_length": 2048,
30
+ "max_num_images": 5,
31
+ "max_topk": 256,
32
+ "n_active_tokens": 2048,
33
+ "neuronxcc_version": "2.21.33363.0+82129205",
34
+ "on_device_sampling": true,
35
+ "optimum_neuron_version": "0.4.6.dev4",
36
+ "output_logits": false,
37
+ "pp_degree": 1,
38
+ "prefill_chunk_size": 0,
39
+ "sequence_length": 2048,
40
+ "speculation_length": 0,
41
+ "start_rank_id": 0,
42
+ "target": "trn1",
43
+ "torch_dtype": "bfloat16",
44
+ "tp_degree": 2
45
+ },
46
+ "text_config": {
47
+ "_sliding_window_pattern": 6,
48
+ "attention_bias": false,
49
+ "attention_dropout": 0.0,
50
+ "attn_logit_softcapping": null,
51
+ "cache_implementation": "hybrid",
52
+ "dtype": "bfloat16",
53
+ "final_logit_softcapping": null,
54
+ "head_dim": 256,
55
+ "hidden_activation": "gelu_pytorch_tanh",
56
+ "hidden_size": 2560,
57
+ "initializer_range": 0.02,
58
+ "intermediate_size": 10240,
59
+ "layer_types": [
60
+ "sliding_attention",
61
+ "sliding_attention",
62
+ "sliding_attention",
63
+ "sliding_attention",
64
+ "sliding_attention",
65
+ "full_attention",
66
+ "sliding_attention",
67
+ "sliding_attention",
68
+ "sliding_attention",
69
+ "sliding_attention",
70
+ "sliding_attention",
71
+ "full_attention",
72
+ "sliding_attention",
73
+ "sliding_attention",
74
+ "sliding_attention",
75
+ "sliding_attention",
76
+ "sliding_attention",
77
+ "full_attention",
78
+ "sliding_attention",
79
+ "sliding_attention",
80
+ "sliding_attention",
81
+ "sliding_attention",
82
+ "sliding_attention",
83
+ "full_attention",
84
+ "sliding_attention",
85
+ "sliding_attention",
86
+ "sliding_attention",
87
+ "sliding_attention",
88
+ "sliding_attention",
89
+ "full_attention",
90
+ "sliding_attention",
91
+ "sliding_attention",
92
+ "sliding_attention",
93
+ "sliding_attention"
94
+ ],
95
+ "max_position_embeddings": 131072,
96
+ "model_type": "gemma3_text",
97
+ "num_attention_heads": 8,
98
+ "num_hidden_layers": 34,
99
+ "num_key_value_heads": 4,
100
+ "query_pre_attn_scalar": 256,
101
+ "rms_norm_eps": 1e-06,
102
+ "rope_local_base_freq": 10000.0,
103
+ "rope_scaling": {
104
+ "factor": 8.0,
105
+ "rope_type": "linear"
106
+ },
107
+ "rope_theta": 1000000.0,
108
+ "sliding_window": 1024,
109
+ "sliding_window_pattern": 6,
110
+ "use_bidirectional_attention": false,
111
+ "use_cache": true,
112
+ "vocab_size": 262208
113
+ },
114
+ "unsloth_fixed": true,
115
+ "vision_config": {
116
+ "attention_dropout": 0.0,
117
+ "dtype": "bfloat16",
118
+ "hidden_act": "gelu_pytorch_tanh",
119
+ "hidden_size": 1152,
120
+ "image_size": 896,
121
+ "intermediate_size": 4304,
122
+ "layer_norm_eps": 1e-06,
123
+ "model_type": "siglip_vision_model",
124
+ "num_attention_heads": 16,
125
+ "num_channels": 3,
126
+ "num_hidden_layers": 27,
127
+ "patch_size": 14,
128
+ "vision_use_head": false
129
+ }
130
+ }
neuronxcc-2.21.33363.0+82129205/MODULE_15503648c89ebbfa6dc4+24129607/model.hlo_module.pb CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fbfd2d97c8bd9f6ae6e1eac3dc6590c9c46e4524a9233efe50ced80fdc586880
3
  size 1048014
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7157d911dfef03e6e18da2ddb64680ffd5e810c49e95cdaaa4bd9762c86dc9dd
3
  size 1048014
neuronxcc-2.21.33363.0+82129205/MODULE_15503648c89ebbfa6dc4+24129607/model.neff CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5197489ec8a4607c85543665a90d5492077cdd6b0c4c3948b510e947ab322668
3
  size 32861184
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3ace5d73842a56402f400cf46152f4835bb9d57d0d7bf923bbbe2d754ea2835
3
  size 32861184
neuronxcc-2.21.33363.0+82129205/MODULE_2a8f4b110978c8fcb167+c4f887dc/model.hlo_module.pb CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7a3aeeb1d9e3ac163b82c3cb7891c3a8df2dc83db68134f2793e5ceb80eaa62e
3
  size 383063
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9634010586f4568593b689759aa82fdd40a182417f810442ba024bb7dc061c4
3
  size 383063
neuronxcc-2.21.33363.0+82129205/MODULE_2a8f4b110978c8fcb167+c4f887dc/model.neff CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:292fc6c5085157358dde40c3145ca531288845ab29ee619901b619dd8b4d211a
3
  size 18699264
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ea1d4e0eab058b08c44ac3dd6fd2d67563702c93d9f548b32e841af90ca4cfe
3
  size 18699264
neuronxcc-2.21.33363.0+82129205/MODULE_2a8f4b110978c8fcb167+c4f887dc/wrapped_neff.hlo CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dfe2c4f10e5f8641aa0b19e4cef09bd15e6cac8b8514e1605688869c76ca7906
3
  size 18932389
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f38e73fceb9b1271458ad24a686f5da7be4a95d988ccb355a1b71b17974a1ee5
3
  size 18932389
neuronxcc-2.21.33363.0+82129205/MODULE_976a4227c74e1e5d858d+a02c3a36/model.neff CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:93693cfc44c1704bb08f92dd01eb408aa1f8eb4888603a61da5ab9431c4fc292
3
  size 6933504
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b2c4833d1ec33472f840a6f70a80043824c9ae71901874543067175e297b314
3
  size 6933504
neuronxcc-2.21.33363.0+82129205/MODULE_976a4227c74e1e5d858d+a02c3a36/wrapped_neff.hlo CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fdbbad0e649a2e58665016e3b6e5daaf73cdcbdcef29dabc386d4ab08aec3b56
3
  size 7164278
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8af64065f8063a2173dfd498b980e1617668f73ed19143fda95122c44633885
3
  size 7164278